Modulation classification for radar pulses in low SNR levels with graph features of ambiguity function

Abstract

Owing to unsatisfactory performance of radar modulation classification in low signal-to-noise ratio (SNR) levels, a method that extracts graph features from ambiguity function is proposed. First, the max energy angle is searched for and the graph slice on this angle is extracted. Second, an optimization algorithm that denoises the extracted graph slice based on derivative constraint is presented, which enhances distinctions among different modulation types for classification. Then, the feature vectors are formed by the max energy angle and symmetrical Holder coefficients of the denoised slice. Finally, fuzzy c-means is implemented for modulation classification. Simulation results indicate that the method presented in the paper adopts higher classification accuracy.